Simulation and Detector Optimization G.Cibinetto, N.Gagliardi, M.Munerato and M.Rotondo XII SuperB General Meeting, Annecy, 17/03/2010
Simulation and Detector Optimization
G.Cibinetto, N.Gagliardi, M.Munerato and M.Rotondo
XII SuperB General Meeting, Annecy, 17/03/2010
Outline
● Detector optimization
● Strategy and code structure
● Multivariate Analysis
● Three configuration analyzed
● Efficiencies and misID distributions (as function of p);
● Impact of noise: first look
● Results
● Outlook
Strategy of the IFR Detector Optimization
● Full simulation (BRUNO) used to generate GHits from single particles
● Magnetic field is off to avoid to implement complex swimmers
● Implement the reconstruction in the IFR starting from GHits collected into standard rootples obtained from BRUNO (BERT hadronic list)
● Sample of single pions and muons are simulated
● To understand the effect of different intrinsic IFR geometries we fire particles on a small portion of the barrel
● 3 configurations are considered, corresponding to different total amount of iron
● The reconstructed quantity are given as input to a Multivariate Classifier and the muon efficiency and pion rejection efficiency are compared
● Specific package (IfrRootCode) has been developed to simulate the electronics and the reconstruction
Reconstruction implementation: IfrRootCode
● Digitization: simulate the detector response-> IFRHits. This step background hits can be added, and detector efficiency can be simulated
● Swimmer and clusterization: tracks from the inned detector (use MC truth) are extrapolated into the IFR. All the IFRHits within a cylinder of 30cm of radius are associated to the tracks
● The clusters are used to make a track object IFRTrack. A fit is performed: all the reco quantity, similar to what we have in BaBar, are computed from IFRTrack.
C13
C14 C2'
λ
IFR Configurations studied
C2' Fe 920mm
C13 Fe 820mm
C14 Fe 620mm
Simulated 500k of single muons and pions for each configuration
Momentum: range from 0 to 5 GeV/c with flat distribution. Fired in a restrincted region of the top-sextant of the barrel
Configurations compared using a BDT as multivariate classification algorithm: 9 variables from IFRTrack
|=|=|========|============|============|=======|=====||2|2| 16 | 24 | 24 | 14 | 10 |
|=|=|========|========|========|========|=======||2|2| 16 | 16 | 16 | 16 | 14 |
|=|=|======|======|======|======|=====||2|2| 12 | 12 | 12 | 12 | 10 |
Measured Interaction Length
Output of the IFR Reconstruction: BDT inputs I
ππ
Interaction LengthMuonsPions
Last Layer
IntLength -ExpectedIntLength
Average Hit Multiplicity
Interaction LengthMuonsPions
Output of the IFR Reconstruction: BDT inputs II
MC-Chi2
zy MC-Chi2
xy
Chi2
xy Chi2
zy
MuonsPions
BDT Output
C2'
C14C13
BDT optimization of S/(S+B) obtained on the full momentum range 0-5 GeV/c
considered
BDT discriminant output
muons
pions
A comparisonwith BaBar is available in the Backupslides
Efficiency and mis-id
● Cut on BDT requiring an average mis-ID=2%
● Muon efficiency and the mis-ID extracted as a funtion of track momentum
● C2' seems the best option
C14
C2'
C13
Muon Efficiency Pion Mis-ID
Further study on the BDT I
● BDT optimization performed in 4 bins of momentum
Further study on the BDT II
● Muons efficiency extracted for each momentum bin requiring a pion mis-ID=2%
C2'
C14
C13
52.9±0.357.0±0.251.0±0.3
93.1±0.175.9±0.195.9±0.1
80.7±0.261.5±0.292.1±0.1
87.3±0.256.0±0.292.1±0.1
Muonsefficiency
Anatomy of the pion mis-ID
● About 50% of the surviving pions is due to decay in fly of pions
● Irreducible background: some handle comes from inner detectors: EMC and DIRC
Pions that decays before the first IFR layer
In RED after the cut on the BDT to keep pion mis-ID at 2%
Pions that does not decay in fly, but survive the cuts
In YELLOW the decay layer number before cuts
Muon momentum from B->D semileptonic decay
theta
pBarrel region
BaBar
SuperB
Using FastSimAverage
momentum
● Momentum distribution in SuperB are different from BaBar due to the change in the boost
Results
● From the study the configuration C2' seems the best option
● At low momentum, the large gaps between active layers make some differences: C14 is better
● Add a layer in a C2' like configuration?
● The pion rejection at low moments can be increased using informations from EMC and DIRC
● In SuperB the muon angular distribution is quite different from BaBar:
● Average muon momentum is lower in the FWD and higher in the BARREL
Noise and realistic detector efficiency
● Add 1.5% of noise distributed uniformly in the detector volume
● Scintillator efficiency = 95%
51.0±0.344.2±0.3
95.9±0.191.2±0.1
92.1±0.188.6±0.1
92.1±0.192.1±0.1
Noise=0%εφφ =100%
Noise=1.5%εφφ =95%
C2' configuration
Summary
● Multivariate optimization (BDT) is an useful tool to compare performances of different IFR configurations
● The study performed so far show C2' is the best option
● Informations from other subdetectors (EMC and DIRC) are not included but these will help to reduce the background (½ of the surviving pions are from decays within the inner detectors)
● Next steps:
● Use realistic distribution for the machine backgrounds: from Full Simulation
● Explore different granularity: the background can make differencies
● Start to study KL ID
● We have 3 fine active layers in the inner region
● The background can be an issue: explore different scintillator size
● Distinguish K interacting in the EMC from K interacting in the EMC-IFR gap and in the IFR volume
BACKUP SLIDES
BACKUP SLIDE
C2'
C14C13
Comparison with theBaBar performanceLimited to the BARREL
Thanks C. Vuosalo
BACKUP SLIDE
From C. Vuosalo
Low momentum